VIVEKANANDHA INSTITUTE OF ENGINEERING & TECHNOLOGY FOR WOMEN

IMAGE PROCESSING

S.P.YAMUNADEVI aarthi_20072007@rediffmail.com (2nd year IT) R.VINOTHINI (2nd year IT) INTRODUCTION
What is an image? A digital representation of a real-world scene. (Graphics is the digital representation of an imaginary scene.) Composed of discrete elements generally called picture element (or pixels for short) Pixels are parameterized by position intensity time In all combinations, these parameters define still images, video, volume data and moving volumes What is image processing? Image processing typically attempts to accomplish one of three things restoring images enhancing images understanding images Restoration takes a corrupted image and attempts to recreate a clean original Enhancement alters an image to makes its meaning clearer to human observers Understanding usually attempts to mimic the human visual system in extracting meaning from an

About image process

X360 Tiff Image Advantages Processing Natural Gas Processing Principals ... Processing ActiveX of distributed 515 x 473 - 12k - gif 584 x 465 - 22k - png ... processing ... 751 x 768 - 120k 435 x 306 - 17k - gif

Processing the money when it comes ... 1297 x 745 - 75k

Food Processing Form Processing Laser polymer processing. Fig. Machines Process, Form ... 412 x 412 - 50k - jpg 360 x 360 - 23k - jpg 413 x 464 - 9k - gif

APPLICATION: IMAGE PROCESSING We apply PM to static black and white images of driving cars (see Figure 2). Each image is divided into square pixels. Each pixel can take on 16 different grey levels represented as integers scaled around zero. Input generation. There is a circular ``input area''. Its diameter is 64 pixel widths. There are 32 code units. For each code unit, there is a ``bias input unit'' with constant activation 1.0, and a circular receptive field of 81 evenly distributed additional input units. The diameter of each receptive field is 20 pixel widths. Receptive fields partly overlap. The positions of code units and receptive fields relative to the input area are fixed. See Figure 3. The rotation of the input area is chosen randomly. Its position is chosen randomly within the boundaries of the image. The activation of an input unit is the average grey level value of the closest pixel and the four adjacent pixels (see Figure 4). Learning: heuristic simplifications. To achieve extreme computational simplicity (and also biological plausibility), we simplify the general method from section 2. Heuristic simplifications are: (1) No error signals are propagated through the predictor input units down into the code network. (2) We focus on semilinear networks as opposed to general nonlinear ones (no

hidden units within predictors and code generating net - see Figure 1). (3) Predictors and code units learn simultaneously (also, each code unit sees only part of the total input). These simplifications make the method local in both space and time -- to change the weight of any predictor connection, we can use the simple delta rule, which needs to know only the current activations of the two connected units, and the current activation of the unit to be predicted: in response to input pattern to , each weight of predictor changes according

where is a positive constant. Likewise, to change the weight of any connection to a code unit, we need to know only the current activations of the two connected units, and the current activation of the corresponding predictor output unit: in response to input pattern changes according to , each weight leading to code unit

where

is a positive constant.

Success rate (measuring information throughput) plotted against number of training pattern presentations (logarithmic scale). Results are shown for various pairs of predictor learning rates . . . and code unit learning rates . .

Figure 8: Bright (dark) circles represent positive (negative) weights. With nearly 200 input units per field and a symmetric arrangement of receptive field centers (essentially, on the circular boundary of each field there are 6 other field centers), the weight patterns generated by the system tend to be either offcenter-on-surround-like (see figure) or on-center-off-surround-like. Measuring information throughput. Unsupervised learning is occasionally switched off. Then the number of pairwise different output patterns in

response to 5000 randomly generated input patterns is determined (the activation of each output unit is taken to be 0 if below 0.05, 1 if above 0.95, and 0.5 otherwise). The success rate is defined by . Clearly, a success rate close to 1.0 implies high information throughput. Medical image processing Medical imaging refers to the techniques and processes used to create images of the human body (or parts thereof) for clinical purposes (medical procedures seeking to reveal, diagnose or examine disease) or medical science (including the study of normal anatomy and function). As a discipline and in its widest sense, it is part of biological imaging and incorporates radiology (in the wider sense), radiological sciences, endoscopy, (medical) thermography, medical photography and microscopy (e.g. for human pathological investigations[1]). Measurement and recording techniques which are not primarily designed to produce images, such as electroencephalography (EEG) and magnetoencephalography (MEG) and others, but which produce data susceptible to be represented as maps (i.e. containing positional information), can be seen as forms of medical imaging. In the clinical context, medical imaging is generally equated to Radiology or "clinical imaging" and the medical practitioner responsible for interpreting (and sometimes acquiring) the images is a radiologist. Diagnostic radiography (see Radiography) designates the technical aspects of medical imaging and in particular the acquisition of medical images. The radiographer or radiologic technologist is usually responsible for acquiring medical images of diagnostic quality, although some radiological interventions are performed by radiologists. As a field of scientific investigation, medical imaging constitutes a subdiscipline of biomedical engineering, medical physics or medicine depending on the context: Research and development in the area of instrumentation, image acquisition (e.g. radiography), modelling and quantification are usually the preserve of biomedical engineering, medical physics and computer science; Research into the application and interpretation of medical images is usually the preserve of radiology and the medical sub-discipline relevant to medical condition or area of medical science (neuroscience, cardiology, psychiatry,

psychology, etc) under investigation. Many of the techniques developed for medical imaging also have scientific and industrial applications

SATELITE IMAGE PROCESSING

Satellites can provide huge amounts of data that in principle could be processed and provide very useful information in areas such as agriculture. Frequent types of analyses of these images are classification (e.g., to identify roads, urban areas, types of cultivation, etc.) and rectification and clustering. The size of the images to be processed by remote sensing end users is typically 20-40 Mbytes per spectral band. An image may include up to 7 bands (or more for multitemporal processing). This analysis is carried out with standard image processing packages. These packages have great image visualisation and manipulation capabilities, but numerically intensive computations may take from 1 to 24 hours on Pentium-based platforms or low end workstations. A typical process iterates on parts of an image to find out the optimal algorithm parameters. Then, the algorithm is applied to the whole image. Many of the providers of experience and services in the area of remote sensing are small companies which would greatly benefit from access to fast processing at a reasonable cost. New image acquisition technologies are providing even higher resolution images, making the problem of slow processing even harder to tackle.

Satellite images, e.g., Landsat-TM Scene (5,500 x 6,000 pixels, 7 bands) and full-frame Landsat Thematic Mapper (7,020 samples, 5,761 rows, 3 bands, 8 bits per pixel), were used to validate the MMIPPS package. Successful results (e.g. expected speed-up factors, better quality results) were obtained. These results were also compared with results when using commercial image processing packages (e.g., ERDAS Imagine). The quality of the MMIPPS results is equal to or better than the quality with the commercial image processing packages.

The project has led to a change in the end users practice by allowing much faster initial studies and evaluation of potential alternatives. The effect is a more dynamic service company that is able to react much faster to customer demands than before the project started. MMIPPS offers the possibility of extending the end users service and product range in areas such as environmental disaster management, natural hazard management, environmental monitoring systems, statistical information for national and European authorities, and urban and infrastructure planning. HPCN technology has been able to restructure industrial practice without the need for new HW/SW. Simply changing the way in which the available infrastructure was used has meant new opportunities for the consultancy companies in this sector.

The package has been installed and is in use at the Geospace, Italeco, and Intecs Sistemi premises. Geospace and Italeco are using MMIPPS software in their new services. Intecs Sistemi has started the internal process of integrating MMIPPS with ArcInfo, with the aim of using this integrated environment in future Intecs Sistemi Remote Sensing and GIS projects.

ADVANTAGE Digital Elevation Model (DEM) is a digital data in which each point represents latitude, longitude and height. A digital elevation model has many applications such as geographic information system (GIS), flight simulations, urban planning, virtual reality and military ones. In these applications, the accuracy of the digital elevation models is important. There are various ways to produce digital elevation models from many types of sources such as airborne images, satellite images, etc. The use of satellite images for digital elevation models generation has the following advantages; 1) A scene covers a larger area. 2) Satellite images are naturally a digital data so that automation can be achieved. 3) Considering that many remote-sensing satellites are being launched, it is becoming easier to get satellite images in hand. However, despite of the advantages stated above, generating digital elevation models from satellite images suffers from several shortcomings -accuracy, coverage and execution time. In Valadd-Pro software, these problems can be overcome by adopting the stereo matching algorithm based on the epipolarity of linear push-broom sensor and the intelligent interpolation scheme modeling

Uses
. In particular, digital image processing is the only practical technology for:

Classification ,Feature extraction ,Pattern recognition ,Projection ,Multi-scale signal analysis Some techniques which are used in digital image processing include: Principal components analysis ,Independent component analysis ,Selforganizing maps, Hidden Markov models, Neural networks, Computer graphics ,Computer vision ,GPGPU, Homomorphic filtering ,Imaging ,Satellite imagery ,Machine vision glossary

References
Sorted alphabetically with respect to first author's family name Wilhelm Burger and Mark J. Burge (2007). Digital Image Processing: An Algorithmic Approach Using Java. Springer. ISBN 1846283795. Rafael C. Gonzalez, Richard E. Woods (1992). Digital Image Processing. ISBN 0-201-50803-6. William K. Pratt (1978). Digital Image Processing. ISBN 0-471-01888-0. John C. Russ (2006). The Image Processing Handbook. ISBN 0849372542. Jean Serra (1982). Image Analysis and Mathematical Morphology. ISBN 0126372403. (1988) Image Analysis and Mathematical Morphology Volume 2: Theoretical Advances. ISBN 0-12-637241-1. Bart M. ter Haar Romeny (2003). Front-End Vision and Multi-Scale Image Analysis. ISBN 1-4020-1507-0.